lkml.org 
[lkml]   [2013]   [Nov]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: perf/tracepoint: another fuzzer generated lockup
On Fri, 8 Nov 2013 23:36:58 +0100
Frederic Weisbecker <fweisbec@gmail.com> wrote:


> [ 237.623178] ---[ end trace 40cda30d05d0ffa6 ]---
> [ 237.627769] perf samples too long (3397569 > 2500), lowering kernel.perf_event_max_sample_rate to 50000
> [ 237.637124] INFO: NMI handler (perf_event_nmi_handler) took too long to run: 444.233 msecs
>
> 444 msecs is huge.

I don't think that's the issue you are looking for. Peter told me
there's a timer race with overflows or something (I don't remember the
details, Peter can clarify it), that causes the second timestamp to be
bogus compared to the first, and this makes the code think that the NMI
handler was running much longer that it actually was.

-- Steve

>
> [ 247.186562] perf samples too long (3371028 > 5000), lowering kernel.perf_event_max_sample_rate to 25000
> [ 257.180241] perf samples too long (3344694 > 10000), lowering kernel.perf_event_max_sample_rate to 13000
> [ 267.173920] perf samples too long (3318566 > 19230), lowering kernel.perf_event_max_sample_rate to 7000
> [ 277.167598] perf samples too long (3292642 > 35714), lowering kernel.perf_event_max_sample_rate to 4000
> [ 287.161277] perf samples too long (3266920 > 62500), lowering kernel.perf_event_max_sample_rate to 2000
> [ 297.154956] perf samples too long (3241400 > 125000), lowering kernel.perf_event_max_sample_rate to 1000



\
 
 \ /
  Last update: 2013-11-09 02:21    [W:0.087 / U:0.492 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site